# Speaker Ask

> The first invitation to a speaker, honest and short, and the follow-ups that add something. Enter the slot, the
> candidate your shortlist picked and why, the honest budget line and the date you need an answer, and get the first
> invitation written to the disclosure list and the word cap through the right channel at the right depth - then the
> follow-up sequence that adds a new angle each touch and the reply to a yes-but. Built on the SkillSafe skill
> @samber/event-speaker-cold-outreach, from samber/dev-event-organizer-skills (MIT), by Samuel Berthe.
> It writes the message, not the search: the shortlist arrives from upstream. Run every invitation past a humanizer
> pass before sending. It never invents a fee, a market rate or a fact about the candidate.

https://speaker-ask.skillsafe.ai/

## What is free
- The sourcing packet checked: ranked candidates, the slot brief, each candidate's why and its evidence, the Hot/Warm/Cold/Skip score, travel, the honest budget line, the published contact route and the invited-review lane; a name without its why goes back to sourcing.
- The pole (a community conference, or a vendor or user conference) and what the invitation can honestly offer; the travel gate - under four weeks between the answer date and the event blocks, four to six weeks must be named in the message.
- The first-touch channel (referrer-relayed introduction, direct email to a published address, published DM, in-person ask; a guessed or scraped address and the candidate's community space deleted), the personalization depth (sourced why, fresh work, slot-brief line, program-shaped ask; the merge-field blast deleted) and the follow-up cadence (three-touch angle, single ask, long runway; the bump deleted), re-ranked per candidate with every move named.
- The disclosure list every first invitation must state, and a pasted draft audited against it: each item found or missing with an excerpt, the five-part structure, the 150-word cap, the subject under 50 characters, the banned vocabulary and phrases, links and attachments.
- The dated follow-up schedule with the new angle each touch carries (a confirmed co-speaker, a program detail, an audience question, a constraint update, a narrowed ask), stopping at the answer date; the reply read as yes, no, yes-but or no answer, each condition mapped to the negotiable list and the authority; the close-the-loop record; a verdict (travel gate blocks, fix before writing, rewrite the draft, ready to send, follow up, answer the yes-but, close the loop); exports: the invitation checklist (.md), the follow-up schedule (.csv and .ics), everything (.json).

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (invite or followup), `event`, `candidate`, `draft` (may be empty), `reply` (may be empty) and `facts`.
- `invite` replies with `summary`, `verdict`, `channel`, `depth`, `subject`, `message`, `disclosures`, `words`, `humanize_note` and `caveats`; the page holds the verdict, the picks, the cap, the disclosure items, the dates, the money words and the banned vocabulary to its own. It is offered until a reply is pasted.
- `followup` replies with `schedule` (each touch's date, channel, angle and message), `yes_but` (each condition with its negotiable flag and answer, the reply to send and the authority note - null unless the reply is a yes-but), `close` (the record and the referral ask), `opt_out_note`, `caveats` and `note`.

## Links
- API tutorial: https://speaker-ask.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/event-speaker-cold-outreach